Missing recorded sentence start (Open Wake Word ESP32-S3-Box-3)

Hello everyone,
I’m looking to get into assist and noticed something unusual:
When using open wake word (on device or home Assistant processed) and Talking like to Siri, so: „Hey Jarvis, Licht im Schlafzimmer ausschalten“ (No Pause between Wake Word and command, speaking normally) then the recorded Debug-WAV file shows clearly that only something like „ht im Schlafzimmer ausschalten“ is captured (Missing Lic at the start of sentence, sometimes a whole word)

Is there any way to improve the recording speed / Delay? Did anyone else notice this „input-lag“?